Advanced Driver Assistance Based on Front-View and Rear-Side-View Scene Analysis

Abstract

This paper proposes a high-performance advanced driver assistance system that analyses front-view driving scenes and rear-side-view scenes. Dense optical flow analysis is calculated for both views to extract motion information. The system performs ego-lane position identification via an effective fuzzy system and indicates if the vehicle is driving on an inner or outer lane. Extracted flow intensities are utilized as the input for deep convolutional neural networks to issue warning events. The front-view event warning system is more responsive to various types of potential approaching dangers because there is no need to detect vehicles first. The rear-side-view scene analysis provides safety check for vehicle doors. Optical flow information and neural networks are also used for rear-side-view scene analysis. The experimental results have shown that the proposed methods can effective detect events or dangerous conditions and help increase the safety of the drivers and road users.